Retention index based approach for simulation of results and application for validation of compound identification in comprehensive two-dimensional gas chromatography

• Published in: Journal of Chromatography A 2022-08, Vol.1679, p.463394, Article 463394
• Main Authors: Kakanopas, Palathip, Janta, Pannipa, Vimolmangkang, Sornkanok, Hermatasia, Friscilla, Kulsing, Chadin
• Format: Article
• Language: English
• Subjects: Comprehensive heartcut; Database simulation approach; High resolution GC; Identified compound validation; Two-dimensional GC
• ISSN: 0021-9673
• DOI: 10.1016/j.chroma.2022.463394

•Retention index based curve fitting approach in GC×GC-MS was developed.•This allows 1tR and 2tR simulation and confirmation of compound identification.•Example simulations were provided and compared with different samples.•2tR correlations were obtained with the R2 of 0.80–0.97 based on 542 compound data.•Out-of-trend correlations indicated incorrectly identified peaks in GC×GC-MS. In this work, first and second dimensional retention index (1I and 2I) based calculation approach is established to simulate peak retention times (1tR and 2tR) of samples for the given sets of volatile compounds in comprehensive two-dimensional gas chromatography-mass spectrometry (GC×GC-MS). For the result without 1tR and 2tR data of alkane references (1tR(n) and 2tR(n)), the following steps were applied: (1) curve fitting based on van den Dool and Kratz relationship in order to simulate 1tR(n) using a training set of volatile compounds in a sample with their experimental 1tR data, and (2) simulation of 2tR(n) at different 1tR(n) to construct their isovolatility curves based on a nonlinear equation with p1-p5 parameters and a constant (within the ranges of -0.0052 to 0.0049, -0.6181 to -0.0230, -26.4775 to -0.2698, 0.0050 to 9.6259, -7.2976 to -3.9524 and 0.9157 to 4.0779, respectively). These parameters were obtained by performing curve fitting according to the experimental 2tR data of the same training set with the least square values ranging from 4.58×10−15 to 32.55. Simulation of 1tR and 2tR of target analytes (1tR,sim and 2tR,sim) with known 1I and 2I were performed using 1tR(n) and the simulated isovolatility curves. All the calculations and curve fittings were carried out by using Solver in Microsoft Excel. The approach was applied to simulate results for 542 compounds in several samples including analysis of saffron (Crocus sativas L.), Boswellia papyrifera, acacia honey and incense powder/smoke, perfume and cannabis either reported from literature or from the experiments in this work using different experimental approaches. These were compared with the experimental data showing correlation with the R2 ranges of 0.98–1.00 and 0.80–0.97 for 1tR and 2tR, respectively. This approach was then applied to propose 6 compounds which may be incorrectly identified based on the differences of >2 times of the standard deviations between 2tR,sim and the experimental 2tR in both residue and leave-one-out analyses.